Pirfenidone modifies hepatic miRNAs expression in a model of MAFLD/NASH

miRNAs are involved in the development of metabolic associated fatty liver disease (MAFLD) and nonalcoholic steatohepatitis (NASH). We aimed to evaluate modifications by prolonged-release pirfenidone (PR-PFD) on key hepatic miRNAs expression in a MAFLD/NASH model. First, male C57BL/6J mice were randomly assigned into groups and fed with conventional diet (CVD) or high fat and carbohydrate diet (HFD) for 16 weeks. At the end of the eighth week, HFD mice were divided in two and only one half was treated with 300 mg/kg/day of PR-PFD mixed with food. Hepatic expression of miRNAs and target genes that participate in inflammation and lipid metabolism was determined by qRT-PCR and transcriptome by microarrays. Increased hepatic expression of miR-21a-5p, miR-34a-5p, miR-122-5p and miR-103-3p in MAFLD/NASH animals was reduced with PR-PFD. Transcriptome analysis showed that 52 genes involved in lipid and collagen biosynthesis and inflammatory response were downregulated in PR-PFD group. The expression of Il1b, Tnfa, Il6, Tgfb1, Col1a1, and Srebf1 were decreased in PR-PFD treated animals. MAFLD/NASH animals compared to CVD group showed modifications in gene metabolic pathways implicated in lipid metabolic process, inflammatory response and insulin resistance; PR-PFD reversed these modifications.

Insulin tolerance test and determination of glucose levels and HOMA. Fasting blood glucose was measured weekly during the study in mouse-tail vein using a clinical glucometer. The Insulin Tolerance Test (ITT) was determined at eighth and sixteenth week. After 4 h-fasting, 0.025 U/mouse of short-acting human recombinant insulin was intraperitoneally administrated and glucose levels were measured at 0, 30, 60 and 90 min after insulin injection. The area under the curve was calculated using the trapezoidal method 30 . Insulin was measured in serum by multiplex detection immunoassay. Insulin resistance (IR) was calculated using the homeostasis model assessment (HOMA)-IR index with the following formula [fasting serum glucose (mg/dL) × fasting serum insulin (μIU/mL)/405] 31 .
Histological analysis of liver. Samples from the three main liver lobes were fixed in 4% paraformaldehyde (0.1 M PBS, pH 7.4). Paraffin embedded tissues were cut into 5 µm sections. Hematoxylin-eosin, Masson's trichrome and Sirius red staining were performed. Masson's and H&E were analyzed by two independent pathologists to quantify inflammation, microvesicular steatosis and macrovesicular steatosis. Microvesicular steatosis was defined by the presence of numerous small lipid vacuoles that do not displace the nucleus; while macrovesicular steatosis was defined by a large lipid vacuole that displaces the nucleus. In addition, periportal and centrilobulillar fibrosis were examined by scoring fibrotic bridges. To evaluate hepatic stellate cells activation, immunoreactivity against alpha-smooth muscle actin (Cell Signaling Technology Inc., Beverly MA, USA) and anti-GFAP (Glial Fibrillary Acidic Protein) (Biocare Medical, Pacheco, CA, USA) was performed using a 1:50 and 1:100 ab dilution respectively. To evaluate inflammatory infiltration in the liver immunoreactivity against CD68 (Biocare Medical, Pacheco, CA, USA) was carried out using a 1:100 ab dilution.
Collagen, alpha-smooth muscle actin, GFAP and CD68 staining were evaluated using Image-Pro software in 30 photographs at 20 × microscopic field magnification (Media Cybernetics, Inc., Bethesda MD, USA).
RT-qPCR. RNA isolation from liver was performed according to Chomczynski and Sacchi modified method 32 . RNA quantity and quality were determined with NanoDrop equipment (Thermo Scientific, USA). 2 µg of total RNA were used for retrotranscription with 240 ng Oligo dT, 0.5 mM dNTPs mix, 10 mM DTT, 2 U of RNAse inhibitor and 200 U M-MLV (Invitrogen, Carlsbad, CA). qPCR reactions were performed on the LightCycler 96 Instrument (Roche Molecular Systems, Pleasanton, CA, USA). All data were run in triplicate, normalized using GAPDH as housekeeping gene and data analysis was performed using the 2 −Δct method 33 . Information about the probes used is shown in Supplemental Table 1. miRNAs extraction and expression. Representative

Results
Pirfenidone improves clinical and biochemical parameters in MAFLD/NASH animal model. All animals gradually increased their weight during the experimental period. As seen in Table 1  www.nature.com/scientificreports/ mice treated with PR-PFD, diminished body weight 18.11% against HFD group (40.33 g ± 2.82 vs 49.25 g ± 2.36; p < 0.05). These data indicated that PFD-treated animals even when fed with HFD, ended with lesser weight gain and final body weight. No statistical differences in liver weight or liver/animal weight ratio were observed in any experimental group (Table 1). Epididymal fat pad showed a significant decrease in PFD group compared to animals in HFD group (1.79 g ± 0.16 vs 2.60 g ± 0.33; p < 0.05) ( Table 1). Also, PFD group displayed a significant decrease in AST and ALT serum levels (p < 0.0001) versus HFD group (Table 1). Triglycerides, cholesterol and VLDL serum levels were higher in HFD group mice against PFD group (p < 0.0001) ( Table 1). Animals treated with pirfenidone had a slight decrease in serum glucose compared to HFD group. However, mice receiving PFD treatment showed increased insulin sensitivity (AUC determinations), which was statistically significant at the end of week 16, (p < 0.05). Insulin levels were increased in HFD animals compared to CVD group (4326 ± 446.3 vs 2490 ± 321.6; p < 0.05) ( Table 1). Insulin showed no statistical differences between HFD and Pirfenidone group. Insulin resistance was increased in HFD group, as assessed by HOMA-IR, yet an increase in this hormone sensitivity was observed in mice treated with PFD.
miR-34a-5p is usually expressed in the liver, modulating oxidative stress and lipid metabolism. Figure 1C shows its increased expression in HFD group compared to PFD administered animals (5060 ± 2844 vs 1264 ± 149.0; p = 0.12). Noteworthy, PFD group levels are similar to those obtained in CVD group (1208 ± 260.4). miR-103-3p has been shown to regulate insulin sensitivity and glucose homeostasis and is highly expressed in the liver of MAFLD patients. Data in Fig. 1D for miR-103-3p indicated an upregulation in HFD animals (150 ± 85.08) and a notable decrease in its expression after PFD treatment (42.14 ± 7.19; p = 0.25). In microarray assays, its target gene Fgf5 was found downregulated (− 1.82); while Cpt1a was found upregulated (0.84) in PFD group. Concurrent with this finding, Cpt1a is upregulated in animals treated with PFD (1.630 ± 0.3047, p < 0.05) compared to the high fat diet group (0.792 ± 0.05).
MAFLD/NASH-related parameters are decreased in liver tissue of animals after pirfenidone administration. Histological analysis of liver of animals fed with HFD group showed significant tissue damage with steatosis, imminent fibrosis, and inflammatory changes predominantly in the periportal area with neutrophils and mononuclear cells. After prolonged-release pirfenidone treatment; as shown in H&E staining, a significant reduction in inflammation foci was achieved (p < 0.001) ( Table 1, Fig. 2). To correlate this data, macrophage recruitment in the liver was evaluated using CD68 marker (Fig. 3A). Quantification of CD68 positive area was increased in HFD animals compared to PFD group. (6.29 ± 1.03vs. 1.57 ± 0.34, p < 0.0001). In particular, macrovesicular steatosis reached substantial reduction in PFD group compared to HFD mice group (Table 1, p < 0.0001); while microvesicular steatosis decrease in PFD group did not reached statistical significance (Fig. 2, Table 1). An imminent fibrosis development was observed, especially in periportal and pericentral areas. Fibrosis assessment was higher in HFD group; while histological improvement was observed in PFD animals (p < 0.01 and p < 0.05 for periportal and pericentral zones, respectively) ( Fig. 2 Sirius red and Masson staining, Table 1). Likewise, a significant decrease in the number of fibrotic bridges was observed after administration of PFD versus HFD group (p < 0.05; Fig. 2 Masson staining, Table 1). To corroborate these facts, Sirius red collagen staining revealed lesser reactivity in PDF treated animals compared to non-treated HFD group (p < 0.05, Fig. 2B).

Hepatic stellate cells activation in a MAFLD/NASH model was reversed in PFD treated animals.
In order to monitor changes in HSC activation, immunohistochemistry for alpha-SMA was realized. As indicated in Fig. 2C, images were analyzed and percentage of area with immunoreactivity was calculated. HFD group showed a 2.28 ± 0.36% predominantly in the periportal area; whereas PFD group only had a 0.275 ± 0.055% (p < 0.01) indicating reduction of HSC activation in treated animals. This data also correlates with lesser fibrosis and collagen staining in this group. To evaluate if pirfenidone decrease α-SMA positive cells is due to reversal into the HSC quiescent phenotype, we performed GFAP immunohistochemistry (Fig. 3B). GFAP positive area was increased in PFD animals compared to HFD group. ( 28 . Fold reduction is indicated in last column of Table 2 when comparing PFD vs HFD groups. Microarrays analysis. Double-channel chips were analyzed to compare 22,000 genes in mice genomes in the three experimental groups. As shown in Fig. 4A the comparison between CVD animals and HFD mice showed that 82 genes were overexpressed. Pathways associated were: lipid metabolic process (51 genes: Acox2, Acbd3, Ch25h, Cpt1a), cholesterol metabolic process (9 genes: Ch25h, Vldlr, Apoa4, Mvk), negative regulation of IL6 production (7 genes: Irak3), lipid homeostasis (7 genes: Apoa4, Acadm) and negative regulation of IFNgamma production ( Irs1, Igf1, Irs3) and Ampk signaling pathway (6 genes: Akt1s1) as indicated in Fig. 4B.   Ch25h, Srebf2, Hacd4), negative regulation of Nfkb transcription factor activity (10 genes: Cmklr1, Nfkbia, Pias4), Finally, oxidative stress-induced gene expression via NRF2 (5 genes: Hmox2, Prkca) and hepatic stellate cell activation (4 genes: Dgat1) (Fig. 4D) were suppressed as well.

Discussion
Several studies have shown that specific miRNAs play a key role in the progression of metabolic associated fatty liver disease (MAFLD), mostly using high-fat diet animal models. Therefore, modulation of miRNA expression could be a potential therapeutic target for the treatment of this disease because miRNAs regulate lipid synthesis, glucose and fatty acid catabolism, inflammation, proliferation, apoptosis and necrosis; all processes epigenetically  Table 2. Gene expression of proinflammatory cytokines and Collagen Type I. Data are expressed in mean ± SEM. CVD conventional diet, HFD high fat and high carbohydrate diet, PFD high fat diet + prolonged release pirfenidone. *p < 0.05, ****p < 0.0001 compared to CVD group and *p < 0.05, **p < 0.01, ****p < 0.0001 as compared with HFD group. a Significance difference versus CVD group (*p < 0.05, **p < 0.01, ****p < 0.0001). b Significance difference versus HFD group (*p ≤ 0.05, **p < 0.01, ****p < 0.0001).  15,23 . A comprehensive literature search was carried out to select some of the most representative miRNAs for each of the key processes found to be dysregulated in non-alcoholic steatohepatitis, which are inflammation, fibrosis, steatosis and insulin resistance. Therefore, alterations on miRNAs target genes involved in liver energy metabolism, inflammation, cell regeneration and fibrogenic signaling; driving the progression from MAFLD to NASH were considered. Our study is the first to show that miR-21a-5p, miR-122-5p, miR-34a-5p and miR-103-3p expression and numerous of their target genes like, Srebf1, Tgfb1, Fasn and Cpt1a are modified by pirfenidone treatment in a MAFLD/NASH model as part of an improvement in molecular, histopathological and biochemical parameters. miR-21a-5p has been extensively studied in liver diseases, as well as its main target genes. miR-21a-5p hepatic expression is increased in animal models and in patients with MAFLD/NASH [38][39][40] . Similarly, we found a significant increased expression of miR-21a-5p induced by high-fat diet in C57BL/6J mice. miR-21a-5p is consider a profibrogenic miRNA by its effect on Smad7 and also participates in the accumulation of liver lipids by interacting with various factors, such as sterol regulatory element binding transcription factor 1 (Srebf1), 3-hydroxy-3-methylglutaryl-coenzyme A reductase (Hmgcr) and protein binding to fatty acids 7 (Fabp7) [41][42][43] . We found that miR-21a-5p expression was decreased in PFD group, as well as, Srebf1 and Tgfb1. miR-21a-5p hepatic diminution has been associated with improved glucose tolerance and insulin sensitivity, in addition to prevent hepatic steatosis and fatty acid absorption 44 . Tgfb signaling is crucial for fibrogenesis as has been convincingly demonstrated in numerous studies 45,46 . In our study, Tgfb1 decrease in PFD group associates with decreased expression of Col1a1. These results correlated with histological analysis displaying a reduction in the number of fibrotic bridges, periportal and pericentral fibrosis and percentage of collagen stained area in PFD group. We also found a faded alpha-SMA positive area, indicating a contracted activation of HSCs due to PFD treatment; which correlates with an increase in GFAP (Glial fibrillary acidic protein) quiescent marker in HSCs in these group. As stated by Zisser et al. and others, quiescent HSCs are characterized by the cytoplasmic storage of vitamin A and markers include, PPARg, GFAP, and BAMBI [47][48][49] . These observed effects are caused by PFD; a drug widely known for reducing fibrosis and collagen deposition in various organs and diseases, including MAFLD models 28,50,51 . Nuclear receptor peroxisome proliferation-activated receptor alpha (Ppara) is target gene of miR-21a-5p. A study has shown that miR-21a-5p is increased in NASH patients, resulting in diminished expression of Ppara 52 . In this way, miR-21a-5p contributes to cell injury, inflammation, and fibrosis, through its inhibition of the PPARA signaling pathway 23 . Also, Sandoval-Rodriguez et al. had previously reported that protein expression of Ppara is diminished in a mouse MAFLD/NASH model and restored to similar levels than CVD in animals treated with PFD 51 . miR-122-5p is one of the most abundant miRNAs in the liver, it constitutes 70% of all hepatic miRNAs. It has been reported that miR-122-5p is overexpressed in liver tissue of C57BL/6 mice with MAFLD, and in HepG2 and Huh-7 cells exposed to FFA. These in vitro models present an excess of lipid accumulation and triglycerides secretion, decreased expression of Sirt1 and genes related to lipogenesis 53 . Similarly, in the present study, we found that miR-122-5p expression was remarkably elevated in HFD mice as determined by qRT-PCR. Conversely, former studies have shown that miR-122-5p is decreased in MAFLD models induced by HFD. Also, reduced levels of miR-122-5p were identified in liver biopsies of patients with NASH, at the most severe stage of the MAFLD spectrum, compared to control group 39 . A possible explanation for these divergent results is the diverse animals' strains and variety of conditions of MAFLD induction, as well as the different co-morbidities and severity in patients with MAFLD. miR-122-5p has been reported to be expressed primarily in hepatocytes and to target multiple enzymes in lipid metabolism, including fatty acid synthase (Fasn), 3-hydroxy-3-methyl-glutarylcoenzyme A reductase (Hmgcr), and Srebf1 and Srebf2 54 . We found that miR-122-5p reduced expression in PFD group correlates with Fasn mRNA reduced expression in microarray. Fasn catalyzes the last step in fatty acid biosynthesis and therefore it is an important determinant of maximal liver ability to generate fatty acids by novo lipogenesis 14 . www.nature.com/scientificreports/ Recently, miR-34a-5p has been shown to be specifically modulated in liver diseases. Circulating miR-34a-5p levels are high in patients with MAFLD and in animal models of steatosis 55 . Similarly, our data showed a higher hepatic miR-34a-5p expression, while PFD administration attenuates its induction. Also, a decrease in the expression of miR-34a-5p and an increase in its target genes Sirt1, Ppara, and Insig2 has been reported when green tea was used to protect against MAFLD development in a rodent model 56 . Hnf4a is a target gene of miR-34a-5p known to modulate the regulatory elements of promoters and enhancers of genes involved in the metabolism of cholesterol, fatty acids and glucose. In microarray analysis, our data showed increased Hnf4a expression by PFD treatment. Specifically in the liver, Hnf4a activates hepatic gluconeogenesis and regulates expression of several genes, including apolipoproteins 57 .

CVD HFD PFD Fold change versus HFD Regulation
miR-103-3p has been shown to regulate insulin sensitivity and glucose homeostasis, previous studies have found that miR-103-3p expression was increased in the liver of patients with MAFLD; as well as, in vitro and in vivo experimental models of MAFLD/NASH [58][59][60] . In our study, miR-103-3p expression was also detected increased due to high-fat diet induction of NASH. Silencing miR-103-3p improved insulin resistance. In contrast, the gain of function of miR-103-3p in the liver was sufficient to induce insulin resistance 21 . Other studies confirmed that high expression of miR-103-3p led to insulin resistance by decreasing the expression of caveolin-1, which is a direct target of miR-103-3p and a critical regulator of insulin receptor 21,22 . These data correlate with the effects seen after PFD treatment, since ITT was improved and miR-103 expression was reduced. In accordance with the miRWalk database 2.0, Srebf1 is a predicted miR-103-3p target gene. Srebf1 is one of the main regulators of de novo lipogenesis in MAFLD, and Srebf1 overexpression contributed predominantly to lipids accumulation 61 . In resveratrol-treated obese rats, a significant decreased expression of miR-103-3p was reported, as well as, a decline in Srebf1 protein expression 60 . According to these facts, in PFD animals we observed the same pattern regarding these molecules. In agreement to the miRWalk database 2.0, carnitine palmitoyltransferase 1a (Cpt1a) is a predicted target gene for miRNA-103-3p. Cpt1a is a key enzyme involved in mitochondrial β-oxidation, catalyzes the transfer of acyl groups from fatty acyl-CoAs to carnitine in the mitochondrial membrane for the translocation into the intermembrane space 62,63 . Cpt1a expression is decreased in our MAFLD model induced by high-fat diet, while increased in PFD group.
Key cellular processes are found to be altered in the development of non-alcoholic steatohepatitis (NASH). Using microarray analyses, we examined genes involved in lipid metabolism, inflammation, oxidative stress, fibrosis, insulin resistance, and how these pathways or cellular metabolism were modified in NASH and after administration of prolonged-release pirfenidone.
When comparing PFD group to HFD group, we found a decreased expression in genes involved in hepatic stellate cells activation, collagen metabolic processes and transforming growth factor β receptor binding, events involved in fibrosis progression, which is coherent with the well-known antifibrotic effect of pirfenidone in diverse fibrotic diseases 27,28,64,65 .
Particularly in NASH related to MAFLD, recently Komiya et al. and Sandoval-Rodriguez et al. found that pirfenidone improves liver fibrosis in mice models 50,51 . Gutierrez-Cuevas et al. also found in a model of MAFLD induced by High fat diet that PFD showed beneficial effects in cardiac fibrosis 66 .
Insulin resistance is a characteristic feature of MAFLD that contributes to its pathogenesis. Under conditions of insulin resistance, abnormally high levels of insulin are required to metabolize glucose and inhibit hepatic glucose production effectively due to reduced insulin sensitivity in peripheral tissues. Insulin resistance stimulates the pancreas to increase insulin secretion from the portal vein, leading to higher levels of insulin in the liver than in the periphery. High concentrations of hepatic glucose and plasmatic insulin are recognized as biomarkers of hepatic insulin resistance. Elevated fasting glucose levels are due to hepatic insulin resistance, while elevated concentrations of free fatty acids occur 67 . Microarrays showed a decrease expression in genes that participate in insulin resistance and in insulin secretion in the PFD group compared to HF group, which indicates that hyperinsulinemia might be caused by a high fat diet.
Activation of innate response of the immune system during NASH occurs mainly due to excessive accumulation of lipids that leads to hepatocytes damage, activating an inflammatory response that worsens the progress of liver disease. Inflammation is characterized by Tnfa, Il1b, as well as reactive oxygen species production 68 . Our data indicates a decrease expression in genes involved in the innate response of the immune system in PFD mice versus HFD group. Extensive experimental and clinical data support a central role for macrophages in the development and progression of NASH. Liver-resident Kupffer cells initiate inflammation and help recruit blood-derived monocytes 69 . Macrophage polarization to M1 phenotype (proinflammatory) contribute to NASH progression 70 . Here, we found a significant reduction in immunoreactivity of CD68 marker in PFD treated animals. These data suggest that pirfenidone is able to reduce inflammatory response activation, correlating with the already reported anti-inflammatory properties of PFD 28,71 .
Nfkb is a transcription factor involved in innate and adaptive immune responses, as well as in a number of pathological processes, such as inflammation. Under normal conditions, Nfkb is sequestered in the cytoplasm and binds to IKB proteins, inhibiting the nuclear localization of NFKB. NFKB activation is normally moderate, whereas, under conditions of insulin resistance, its expression in the liver is greatly increased. Nfkb translocation to the nucleus leads to upregulation of target genes encoding inflammatory mediators, such as Tnfa and Il6 72 . We found a decrease expression of genes related to the inflammatory response in PFD mice compared to the HFP group, which could be linked to a decrease in the negative regulation of Nfkb activity.
We also found a decreased expression in genes that participate in the metabolic process of lipids and fatty acids, mainly of genes that participate in lipogenesis in PFD animals compared to HFD group. On the other hand, microarrays showed an increase in the expression of genes related with activity of the cholesterol transporter, the transport of lipids and type 1 fatty acid desaturase. Previous studies have reported changes in the expression of liver desaturases during the development of MAFLD/NASH, it was found that a significant decrease in the activity of Fads1 in the progression to NASH 63  www.nature.com/scientificreports/ free cholesterol has been shown to be an outstanding feature in MAFLD, which correlates with the histological severity of the disease. Additionally, epidemiological studies have identified dietary cholesterol intake as a factor related to the risk and severity of MAFLD. Transport of lipids is important, since an excessive flow of free fatty acids into the liver can cause NASH. Diet is an important component in the development of MAFLD, since 15% of triglycerides in the liver have the diet as their source 73 . Fatty acids in the diet are involved in liver lipogenesis and could play a dual role in the pathogenesis of liver steatosis, since they are involved in its development and in its prevention or reversal depending of the amount of omega-fatty acids. Lipotoxicity, arising from hepatic fat excess, leads to mitochondrial dysfunction associated with an elevated ability to oxidize fatty acids, resulting in the production of reactive oxygen species (ROS). In transcriptome analysis, a decreased expression in genes involved in the response to oxidative stress was present in PFD group compared to HFD group. MAFLD causes overexpression of antioxidant agents, but the unbalanced production of ROS promotes oxidative stress. Pirfenidone has been reported to have antioxidant capacity 74,75 and this effect correlates with expected benefits in this MALFD model. Oxidative stress can induce hepatocellular damage by inhibiting enzymes involved in mitochondrial respiratory chain and inactivation of glyceraldehyde-3-phosphate dehydrogenase 76 . Furthermore, ROS cause lipid peroxidation and cytokine production, which contributes to hepatocellular injury and fibrosis, promoting the progression from simple steatosis to nonalcoholic steatohepatitis 68 . Oxidative stress in MAFLD induces hepatic stellate cell activation, the most important producer of extracellular matrix 77,78 . In summary, increased expression of miR-21a-5p, miR-34a-5p, miR-122-5p and miR-103-3p in this MAFLD/NASH model was reversed with prolonged-release pirfenidone. MAFLD/NASH group compared to conventional diet control revealed modifications in gene metabolic pathways implicated in lipid metabolic process, inflammatory response, antioxidant activity and insulin resistance; PR-PFD seems to reverse these modifications.